Neural Representation of Perspective Shapes and the Effect of Attention: Evidence from fMRI and MEG

Introduction

We always perceive objects from a specific perspective. In many cases, the shape we rely on when observing an object, the perspective shape, differs from its objective shape, the physical shape in the real world. For instance, a coin rotated 45 degrees appears elliptical or olive-shaped from the perspective shape, while its objective shape is circular.

Primates are capable of reliably identifying objects even under conditions of different perspectives, positions, and sizes. Numerous studies have shown that primates maintain a robust neural representation of objective shapes despite changes in perspective, position, or size. However, the neural representation of perspective shapes has attracted far less attention compared to objective shapes.

There is controversy over the traditional view of the existence of perspective shapes. Some believe that perspective shapes are directly visible, whereas others think they are constructed through imagination. There remains a breadth of divergence in understanding perspective shapes. Some advocate that the perception of perspective shapes involves cognitive participation and in-depth interpretation, while others deny its role in high-level cognitive processing, considering perspective changes as purely sensory experiences.

Recent studies have reignited this topic. They found that participants took longer to respond under perspective shape matching interference conditions using a forced-choice paradigm, providing evidence for the existence of perspective shape perception. Building on this work, we aim to explore the neural representation of perspective shapes and its modulation by attention.

Source

This article was co-authored by Yi Lin, Yung-Yi Hsu, Tony Cheng, Pin-Cheng Hsiung, Chen-Wei Wu, and Po-Jang Hsieh from different countries and research institutions, including the National Cheng Kung University and Academia Sinica. It was published in the journal “Cortex” (Volume 176), with the digital object identifier 10.1016/j.cortex.2024.04.003. The article was accepted on April 5, 2024, and published online on April 26, 2024. This is an open-access article.

Experimental Methods

To address the issues mentioned, we employed functional magnetic resonance imaging (fMRI), magnetoencephalography (MEG), and multivariate decoding techniques to study the spatiotemporal neural representation of perspective shapes. Participants observed rotated objects, and by measuring their brain activity, we analyzed the neural representation of basic shapes and perspective shapes.

Sample Selection and Experimental Tasks

A total of 22 right-handed Mandarin speakers participated in the fMRI experiment, and 21 completed the MEG task. In the experiment, participants performed object shape and perspective shape recognition tasks, each repeated 5 times, for a total of 10 trials.

Stimulus Materials

The stimuli used in the experiment were coin-shaped objects created by Morales et al., which generated different perspective shapes (e.g., circles, rotating circles, ellipses, large ellipses, and small circles) through rotation. These stimuli were used for shape discrimination tasks.

Data Collection and Preprocessing

MRI data were collected using a 3T scanner, and MEG data were collected in a magnetically shielded room using a 306-channel whole-head system. Preprocessing of MRI data included steps such as time-series correction, motion correction, and spatial normalization. MEG data were preprocessed using MaxFilter software to remove external noise, apply filters, perform signal space separation, and clean artifacts.

Data Analysis

Analysis of fMRI data was carried out using SPM12 software, including first-level analysis and pattern decoding analysis. MEG data’s pattern decoding analysis was performed using MATLAB and support vector machines (SVM). By training different shape classifiers (e.g., circles vs. ellipses, circles vs. large ellipses, and small circles vs. ellipses), we compared the degree to which neural representations tended toward perspective shapes.

Experimental Results

fMRI and Perspective Shape Representation

fMRI pattern decoding results showed that several brain regions (including bilateral occipital, occipital-temporal cortices, and occipital lobes) exhibited significant differences in neural representation between circles and ellipses. In object shape discrimination tasks, some brain regions (such as the left occipital pole and right lateral occipital cortex) showed a trend of neural representation inclination toward perspective shapes.

MEG and Temporal Dimension

MEG pattern decoding results indicated that during object shape discrimination tasks, from around 100ms to 200ms, the neural representation of rotating circles in the occipital region tended toward perspective shapes. In the perspective shape discrimination tasks, a longer inclination effect was observed, starting at 100ms and continuing until 200ms. Compared to the object shape discrimination task, in the perspective shape discrimination task, the temporal lobe also exhibited an inclination effect after approximately 300ms.

Conclusion

This study demonstrates that the neural representation of perspective shapes exists in both low-level and high-level visual brain regions and that these representations are modulated by attention. We found that regardless of task type, the lateral occipital cortex plays a significant role in representing perspective shapes of objects. Low-level visual regions (like the occipital pole) reflect complete shape information processing, supporting the “back-projection” hypothesis. The modulation of neural representation by attention is particularly significant in higher-level visual processing regions, suggesting that visual perception is significantly influenced by goal-directed attention. In the perspective shape discrimination task, the left middle frontal gyrus, as the only frontal region showing a trend, validates the semantic processing and decision-making processes of perspective shapes.

This study reveals the spatiotemporal characteristics of the neural representation of perspective shapes, confirming their existence and significant modulation by attention. Future research should further explore these mechanisms to deepen the understanding of perspective shape perception and attentional regulation.